Extract Format Information From PDF in C#

Question

I am required to work out a program that can analyse and understand contextual and semantic relations of the content in some PDFs with a particular structure and format.

Below is a sample, which shows a piece of content of such PDF:

So I need a PDF reading library which can extract not only the text, but also the metedata in PDF, such as font size, font style(bold, itali), background color, table and its children elements, table cell background color, check box, element position and so on.

Is there any free library of .NET can do the job? Thanks so much.

PS: I am aware of this post: Extract Data from .PDF files, but the libraries' capbilities are not elaborated.

Answer 1

I don't have a quick answer, but I've spent the last two weeks solving this exact problem, with success. I used Apache PDFBox, which extracts PDF text to TextPositions. These TextPositions contain information about each character in the text (position, bold, italic, font, etc). I used this information to set up bounding boxes for all of the table elements and decifer things like text-alignment, column membership, etc, and then recreate the PDF page and it's tables in Excel, in just under 1000 lines of code.

I did not have to extract graphic elements like checkboxes, but Apache PDFBox does extract to COSStreams, and graphic and form elements can likely be parsed from those streams - I'm not there yet. My code would be able to rebuild the table you showed and would only be missing the checkboxes and background colors.

I've searched for a simpler solution than mine and came up short, it seems there's no easy way to do this.

EDIT: If this hasn't dissuaded you, I can show you how to begin. First, extend either PDFTextStripper or PDFTextStripperByArea. This gives you access to the TextPositions via the processTextPosition override - the following code shows how I transformed TextPositions into my own custom class TextChar. I then use relative textpositions to work out rudimentary contextual information:

public class PDFStripper : PDFTextStripper
    {
        private List<TextChar>[] tcPages;

        public PDFStripper(java.util.List pages)
        {
            int pagecount = pages.size();
            tcPages = new List<TextChar>[pagecount+1];
            base.processPages(pages);
        }

        protected override void processTextPosition(TextPosition tp)
        {
            PDGraphicsState gs = getGraphicsState();
            TextChar tc = BuildTextChar(tp, gs);
            int currentPageNo = getCurrentPageNo();
            if (tcPages.ElementAtOrDefault(currentPageNo) == null)
            {
                tcPages[currentPageNo] = new List<TextChar>();
            }
            tcPages[currentPageNo].Add(tc);
        }

        private static TextChar BuildTextChar(TextPosition tp, PDGraphicsState gstate)
        {
            TextChar tc = new TextChar();
            tc.Char = tp.getCharacter()[0];

            float h = (float)Math.Floor(tp.getHeightDir());
            tc.Box = new RectangleF
            (
                tp.getXDirAdj(),
                (float)Math.Round(tp.getYDirAdj(), 0, MidpointRounding.ToEven) - h, // adjusted Y to top
                tp.getWidthDirAdj(),
                h
            );

            tc.Direction = tp.getDir();
            tc.SpaceWidth = tp.getWidthOfSpace();

            tc.Font = tp.getFont().getBaseFont();
            tc.FontSize = tp.getFontSizeInPt();

            try
            {
                int[] flags =     
                     GetBits(tp.getFont().getFontDescriptor().getFlags());
                tc.IsBold = findBold(tp, flags, gstate);
                tc.IsItalic = findItalics(tp, flags);
            }
            catch { }

            return tc;
        }

        protected override void writePage() { return; } //prevents exception
    }

Answer 2

Add this NuGet package https://www.nuget.org/packages/TikaOnDotNet/ . It's the dotnet version of apache tika

then do this:

var extracted = new TikaOnDotNet.TextExtractor().Extract("file.pdf");
var text = extracted.Text;
var metaData = extracted.Metadata;

Good luck buddy :)